1. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 09 Issue: 12 | Dec 2022 www.irjet.net p-ISSN: 2395-0072
© 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 1391
Advance Intelligent Video Surveillance System Using OpenCV
ASHOK KUMAR BANDI1, LAXMAN NANGUNURI2, SUDHARSHAN GUJJERI3, ANOOPAMA K4
4Assistant Professor, Department of Computer Science and Engineering, SNIST, Hyderabad-501301
1,2,3B. Tech Scholars, Department of Computer Science and Engineering, SNIST, Hyderabad-501301
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract
In today's culture, when everyone wants to keep their assets
safe and secure, video surveillance for keeping an eye on a
certain area, such as hospitals, institutions, public parks, and
buildings, has become necessary. Crime has increased as a
result of the growing urban population. Videosurveillance has
significantly impacted citizens. Themostpopulartechnologyis
closed-circuit television (CCTV), but it is more expensive,
consumes more power, and requires morestorage. Wecreated
a cutting-edge intelligent video surveillance system for
settings with erratic human presence to address this problem.
In such circumstances, it is not required to have cameras
trained on the area. This uses up the electricity and the
footage storage space that is needed. Machine Learning tool
called It uses a software library to implement this system. The
proposed system works by capturingvideo, processingitframe
by frame, and starting to record when it recognizes human
presence. If the cameras spot any movement, the security
system will turn on. The proposed system collects information
and keeps it in a local database. The video that was recorded
and saved can be used to recognize theintruderandhelptrack
him down. It may be useful in locations with sporadic human
presence, including homes and bank vaults.
Keywords—: Surveillance; Intruder;OpenCV;Face
Recognition
1. Introduction
How computers might be employed to extract profound
insights from digital photos or videos is one of the many
topics covered by the tremendously diverse scientific field
known as computer vision. The objective is to automate
operations that the human visual system is capable of
performing from an engineering standpoint. The study of
computer vision involves techniques for capturing
photographs, processing them, comprehending them as
digital images, and extracting high-dimensional information
from the real world to create information that may be
expressed as numbers or symbols, such as judgements. The
primary objective of a set of programming functions known
as OpenCV is real-time computer vision (Open-source
computer vision). It was initially created by Intel. The
collection is multi-platform and freely accessible to anyone.
Traditional video security systems don't allow for quick
intervention when a crime is being committed. A systemlike
this is also quite expensive and complicated to set up. The
goal of this project is to createanintelligentopen-sourcetool
that can help people. individuals or organizationsinbuilding
a secure, affordable system on their own They will therefore
have total control over their technology, giving them the
ability to secure the settings and modify them to suit their
requirements. In order to preventsomecrimes,itisessential
to secure our homes, places of employment, and other
commercial areas. Typical surveillance equipment is unable
to inform property owners of any criminal activity. The only
thing transmitted and recorded is the feed. As a result, the
owners are unable to prevent a break-in or theft right away.
2. Literature survey
There are many different techniques for facial recognition
available today, but the majority of them have certain
drawbacks. a process for improving face detection accuracy
using the Euclidean distance technique. The model is
equipped with the Euclidean algorithm to recognize faces
from a variety of distances. But this model'sweaknessisthat
it can be fooled with facial photos, and the dataset utilized
isn't big enough to increase detection accuracy.
Motion detection can be used to stop spoofing that employs
simply photographs. Article uses an infrared PIR (Passive
Infrared Sensor) sensor to detect motion in the person,
which allows it to determine whether the person is genuine
or an image. But adding an additional sensor increases the
There is no need for an additional sensor because OpenCV 3
provides libraries with algorithms to detect motioninvideo,
which lowers the cost.
Another study makes use of the LPBH algorithm and motion
detection to identify faces and record their attendance.
However, a person can mark his attendance by wearing
another person's mask.This model isnotentirelyspoofproof
because it cannot identify facial features.
The LPBH method is used in this paper to work with low
resolution images. Although it makes advantage of low-
resolution photos by manipulating pixels, this model still
lacks the spoof-proof features required for security
surveillance.
This device enhances a home's security features,
transforming it into a smart home. [1] The internet of things
(IoT) and the raspberry pi are utilized to comply to all of
these trends. This defense system alerts the subject to the
problem and provides a clean photographofwhat'sgoing on.

2. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 09 Issue: 12 | Dec 2022 www.irjet.net p-ISSN: 2395-0072
© 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 1392
When the temperature in the living room rises, the fireplace
sensor notices it and updates the reputation in the URL that
was given to the customer, activating a buzzer. The strength
of the gas in the air is the only factor used by the fuel sensor
to identify any gas leaks. The automatic door lock is turned
on by a dc motor. All statuses are sent across the Internet of
Things between the sensors andtheuser.Paper3:Raspberry
Pi and OpenCV-based low-cost smart security camera with
night vision capability.
Context: With the help of a Raspberry Pi (RPI) and a PIR
sensor, the system's goal was to develop and construct a
low-cost smart security camera with night vision
capabilities. It is capable of both individual detection and
smoke detection, which may helptoprevent bothcrimesand
fires. PIR sensor-equipped Raspberry Pi (RPI) regulates the
shifting body, supervises the alarm systems, and uses
Bluetooth to send the acquired photographs to the person's
email address. As part of the alarm system, the speech
"intruder" may be heard when an outsider is detected. This
investigation looks at a smartsurveillancemonitoringdevice
built mostly on the raspberry. Video surveillance isessential
for safety in today's society. Commercial spaces must have
high-speed cameras, warehouses, hospitals, schools, and
other unfavorable indoor and outdoor locations. Due to the
expensive RFIDs used in modern solutions, more researchis
required and the security industrybecomesmoreexpensive.
3. Proposed System
OpenCV, an open-source computer vision library, is used to
implement the suggested system.It reducesboththeamount
of power used and the amount of storage needed by only
monitoring data when a human is detected.
Mission's methodology model
1. The camera is initialized in step one.
2. The camera begins to scan the surrounding surroundings.
3. If a person is discovered by the camera, the recording will
start.
4. If the motion is not detected for a few seconds, the
recording may stop.
5. The recorded videos are kept on a local drive.
Architectural Diagram
In addition to showing the relationships, constraints, and
difficulties that exist between each part, the architectural
diagram also shows howthe softwaredevicelooksoverall.
1. Initialization of the camera is done first.
2. The camera's evolution to view the surrounding area
begins.
3. Start up the Haar cascade classifier.
4. Produce grayscale images from the video frames.
5. The recording will begin once the digicam recognizes
someone.
6. The recording may be halted if the guy or woman isn't
always spotted for a few seconds.
7. A neighborhood force stores the video tapes.
4. EXPERIMENTAL METHODOLOGY
Face Detection Algorithms
Eigenfaces (pca set of rules)
In 1991, Turk and Pentland suggested a face recognition
technique based on concepts from linear algebra and
dimensionality reduction. This method was used in a range
of applications at the time, including handwriting
recognition, lip reading, medical pictureanalysis,andothers.
It is computationally less expensive and straightforward to
implement. In 1901, Pearson proposed PCA (Principal
Component Analysis) as a method for reducing the number
of dimensions. It reduces dimensionality using Eigenvalues
and Eigen Vectors.

3. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 09 Issue: 12 | Dec 2022 www.irjet.net p-ISSN: 2395-0072
© 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 1393
Fisher faces
The eigenfaces have been improved by fisher faces. This
approach is predicated on the notion that not every feature
of the face is equally important or useful for facial
recognition. In order to identify someone when we look at
their face, we search for the features thatdifferthemost. The
Fisher faces technique has been shown to be 93 percent
accurate when used in conjunction with the PCA approach
during the pre-processing stage.
The fisher faces set of rules has been reported to be 93%
accurate when used with the PCA technique on the pre-
processing step.
Fig 2: Fisher faces
KERNEL METHODS: PCA AND SVM
PCA
PCA (principal component analysis), when applied in the
face recognition process, tries to reduce the amount of
source data while maintaining the most crucial information.
It generates a number of weighted eigenvectors, which in
turn generate eigenfaces, vast collections ofdifferentimages
of human faces. Each image in the training set is represented
by a linear combinationof eigenfaces.Theseeigenvectors are
derived using PCA from the covariance matrix of a series of
training images. The major components of each image are
calculated (from 5 to 200). The remaining components
reveal subtle distinctions between noise and faces. The
principal component of the unknown image is contrasted
with the major components of all other images as part of the
recognition process.
SVM
Support vector machine (SVM) is a machine learning
technique that uses a two-group classification notion to
distinguish between "faces" and "not-faces." Toclassifyfresh
test data, a model employs a labelled training data set for
each category. Researchers employ both linear and
nonlinear. SVM training models for face recognition.
According to current results, the nonlinear trainingmachine
has a bigger margin and better identification and
classification results.
LBPH Algorithm
A technique for facial identification called LBPH (Local
Binary Patterns Histograms) makes use of LBP. By
thresholding each pixel's neighborhood and treating the
result as a binary integer, the Local Binary Pattern (LBP)
texturing operator assigns labels to individual pixels inside
an image. Additionally, it was foundthatcombiningLBPwith
the histograms of oriented gradients (HOG) descriptor
considerably improves detection performance on specific
datasets. The facial images can be represented by a
straightforward data vector. LBP may be used to identify
faces because it is a visual descriptor.
Pros
1. Increased precision
2. Easy to understand
Cons
1. Nevertheless, there might be a loss of recognition.
2. It takes a long time to receive acknowledgement.
Fig 3: Facial Recognition through LBPH
In their 2001 paper "rapid item popularity with a boosted
cascade of easy features," Paul viola and Michael jones
suggested an efficient object popularity strategy that was
mostly based on the haar function based on cascade
classifiers. It is a system-learning-based technique that
investigates a cascade characteristic using a huge variety of
fantastic and dreadful images. Then it is applied to several
snapshots so that objects may be located. We might
concentrate on face detectionhere.The approachneedsboth
positive (face-related) and negative (non-face-associated)
images to train the classifier. Thenwe'll decidewhatbenefits
we can derive from it. For this, people with attributes like
those listed below are hired. The picture below
demonstrates the type of features including four rectangle
functions, line functions, and edge capabilities. When
recognizing faces, these abilities are utilized.

4. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 09 Issue: 12 | Dec 2022 www.irjet.net p-ISSN: 2395-0072
© 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 1394
Fig 4: Theoretical Face Model
In a photograph, the non-face area is where most of the
individuals are. It is therefore preferabletohaveaneasy way
to tell whether a window is not always a face location. If it
isn't, throw it away immediately to avoidhavingtoprocessit
again. Instead, focus on locations where a face mightappear.
This enables us to evaluate prospective face regions more
thoroughly.
They developed the idea of a cascade of classifierstoachieve
it. The 6000 capabilities might be divided up into stages of
classifiers and used one at a time rather than all at once.
Fig 5: Detecting Facial Features Using Haar
A window should be discarded if the initial check fails. The
final characteristics are not taken into account.
Haar cascadesareincrediblyquick whencomputinghaar-
like functions since they use the necessary pix (also called
summed location tables). The AdaBoost set of rules makes
them fairly precise in their function selection. The most
important role may be their ability to recognize faces in
images, regardless of their size or function.
Gray Scaling Images
The method of grey scaling involves changing a picturefrom
one colour space to another. In contrast to one-bit bi-tonal
black-and-white images, which are black-and-white images
in the context of computer imaging, grayscale images have
more than two colours. There are numerous shades of grey
in between in grayscale photos.
Significance of grey scaling
1. Reducing the dimensions of the photos: Grayscale
images are only one-dimensional and have no
additional parameter for colour channels, but RGB
images have three colour channels and make up a
three-dimensional matrix.
2. Due to dimension reduction, the information
provided to each pixel is reduced.
3. Reduces the complexity of the model: The neural
network's input nodes will be significantly fewer
when less information is provided to each pixel of
the image. As a result, this makes a deep learning
model less difficult.
4. Colour photos are more challenging to visualise
since the same algorithms or processes may not be
able to extract as much information from some
images using grey scaling. The characteristics
needed for extraction becomemuchmoreapparent.
5. To comprehend image processing better.
Fig 6: BGR to Gray Conversion
.

5. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 09 Issue: 12 | Dec 2022 www.irjet.net p-ISSN: 2395-0072
© 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 1395
Fig 7: BGR to Gray Image
5. CONCLUSION
The device makes use of a haar cascade classifier to locate
faces. It is easy to position into motion. This approach is
particularly beneficial in conditions where humans are
abnormal but the place should be monitored. It can
significantly lessen storage use, power consumption, and
maintenance costs. To function properly, the whole device
best required the setup of OpenCV. It is by far a value
effective era that is easy to put in force. This approach may
be used by everybody, and it facilitates lesseningtheamount
of storage space important for the pictures.
6. REFERENCES
[1]. A. Renjith and Aishwarya, “Enhanced home security
Using IoT and raspberry pi," International Research Journal
of Engineering and Technology (IRJET), vol. 4, 2017.
[2]W. F. Abaya, J. Basa, M. Sy, A. C. Abad, and E. P. Dadios,
"Low-cost smart security camera withnightvisioncapability
using raspberry pi and OpenCV," 2014.
[3]M. Pervaiz, Y. Y. Ghadi, M. Gochoo, A. Jalal, S. Kamal, and
D.-S. Kim, “A smart surveillance system for people counting
and tracking using particle flow and modified som,”
Sustainability, 2021.
[4]M. Rashmika, “Motion sensor and face recognition-based
surveillance system using raspberry pi,” International
Journal of Advanced Research in Computer Science, vol. 8,
no. 5, 2017.
[5]A. CobParro, L. Gutierrez, Marron-Romera,GardelVicente,
and Bravo-Munoz, “Smart video surveillance system based
on edge computing,” 2021.
[6]T. Shivprasad, B. Shivani, A. P. Singh, and Deepak,“Survey
paper on smart surveillance system,”International Research
Journal of Engineering and Technology (IRJET), vol. 3,
2016.Fig 6: BGR to Gray Conversion